Estimation of Distribution Algorithm Using Correlation between Binary Elements: A New Binary-Code Metaheuristic

Pholdee, Nantiwat; Bureerat, Sujin

doi:https://doi.org/10.1155/2017/6043109

Mathematical Problems in Engineering

On this page

Abstract Introduction Conclusions and Discussion Conflicts of Interest Acknowledgments References Copyright Related Articles

Research Article | Open Access

Volume 2017 | Article ID 6043109 | https://doi.org/10.1155/2017/6043109

Estimation of Distribution Algorithm Using Correlation between Binary Elements: A New Binary-Code Metaheuristic

Nantiwat Pholdee¹and Sujin Bureerat¹

Academic Editor: Benjamin Ivorra

Received29 Apr 2017

Revised28 Jul 2017

Accepted02 Aug 2017

Published13 Sept 2017

Abstract

A new metaheuristic called estimation of distribution algorithm using correlation between binary elements (EDACE) is proposed. The method searches for optima using a binary string to represent a design solution. A matrix for correlation between binary elements of a design solution is used to represent a binary population. Optimisation search is achieved by iteratively updating such a matrix. The performance assessment is conducted by comparing the new algorithm with existing binary-code metaheuristics including a genetic algorithm, a univariate marginal distribution algorithm, population-based incremental learning, binary particle swarm optimisation, and binary simulated annealing by using the test problems of CEC2015 competition and one real-world application which is an optimal flight control problem. The comparative results show that the new algorithm is competitive with other established binary-code metaheuristics.

1. Introduction

Nowadays in the economic-competitive world, optimisation has become increasingly popular for real applications as it is a powerful mathematical tool for solving a wide range of engineering design types. Once an optimisation problem is posed, one of the most important elements in the optimisation process is an optimisation method or an optimiser used to find the optimum solution. Optimisers can be categorised as the methods with and without using function derivatives. The former are traditionally called mathematical programming or gradient-based optimisers whereas the latter have various subcategories. One of them is a metaheuristic (MH). The term metaheuristics can cover nature-inspired optimisers [1–10], swarm intelligent algorithms [11–20], and evolutionary algorithms [21–24]. Most of them are based on using a set of design solutions, often called a population, for searching an optimum. The main operator usually consists of the reproduction and selection stages. The advantages of such an optimiser are simplicity to use, global optimisation capability, and flexibility to apply as it is derivative-free. However, it still has a slow convergence rate and search consistency. These issues have made researchers and engineers around the globe investigate how to improve the search performance of MHs.

A genetic algorithm (GA) [21] is probably the best known MH while other popular methods are differential evolution (DE) [22] and particle swarm optimisation (PSO) [17]. Among MH algorithms, they can be categorised as the methods using real, binary, or integer codes. The mix of those types of design variables and some other types can also be made. This makes MHs considerably appealing for use with real-world applications particularly for those design problems that function derivatives are not available or impossible to calculate. Most MHs are based on continuous design variables or real codes. For single objective optimisation, there have been numerous real-code MHs being developed. At the early stage, methods like evolutionary programming [25] and evolution strategies [26] were proposed. Then, DE and PSO were introduced. Until recently, there have been probably over a hundred new real-code MHs in the literature. Some recent algorithms include, for example, a sine-cosine algorithm [27], a grey wolf optimiser [20], teaching-learning-based optimisation [2], and Jaya algorithm [28]. Meanwhile, powerful existing algorithms such as PSO and DE have been upgraded by integrating into them some types of self-adaptive schemes, for example, adaptive differential evolution with optional external archive (JADE) [29], Success-History Based Parameter Adaptation for Differential Evolution (SHADE) [30], SHADE Using Linear Population Size Reduction (LSHADE) [31], and adaptive PSO [32–34]. MHs are even more popular when they can be used to find a Pareto front of a multiobjective optimisation problem within one optimisation run. Such a type of algorithm is usually called multiobjective evolutionary algorithms (MOEAs) where some of the best known algorithms are nondominated sorting genetic algorithm (NSGA-I, NSGA-II, and NSGA-III) [35–37], multiobjective particle swarm optimisation [38], strength Pareto evolutionary algorithm [39], multiobjective grey wolf optimisation [40], multiobjective teaching-learning-based optimisation [41], multiobjective evolutionary algorithm based on decomposition [42], multiobjective ant colony optimisation [43], multiobjective differential evolution [44], and so forth. One of the most challenging issues in MHs is to improve their ability for tackling many-objective optimisation (a problem with more than three objectives). Some recently proposed algorithms are knee point-driven evolutionary algorithm [45], an improved two-archive algorithm [46], preference-inspired coevolutionary algorithms [47], and so forth.

In practice, GA a metaheuristic using binary strings is arguably the most used method as it is included in engineering software such as MATLAB. Apart from GA, other MHs using a binary string representing a design solution include a univariate marginal distribution algorithm (UMDA) [48], population-based incremental learning (PBIL) [24], binary particle swarm optimisation (BPSO) [49], binary simulated annealing (BSA) [50], binary artificial bee colony algorithm based on genetic operator (GBABC) [51], binary quantum-inspired gravitational search algorithm (BQIGSA) [52], and self-adaptive binary variant of a differential evolution algorithm (SabDE) [53]. With the popularity of GA, a binary-code MH has been rarely developed and proposed while its real-code counterparts have over a hundred different search concepts reported in the literature. That means there are possible more than a thousand real-code MH algorithms being published. It should be noted that real-code MHs can be modified to solve binary-code optimisation by means of binarisation [54].

This paper is therefore devoted to the further development of a binary-code metaheuristic. The method is called estimation of distribution algorithm using correlation between binary elements (EDACE). Performance assessment is made by comparing the proposed optimiser with GA, UMDA, BPSO, BSA, and PBIL by using the CEC2015 test problems. Also, the real-world optimal flight control is used for the assessment. The comparative results are obtained and discussed. It is shown that EDACE is among the top performers.

2. Proposed Method

The simplest but efficient estimation of distribution algorithm is probably population-based incremental learning (PBIL). Another MH that uses a similar concept is UMDA. Unlike GA which uses a matrix containing the whole binary solutions during the search, PBIL uses the so-called probability vector to represent a binary population. During an optimisation process, the probability vector is updated iteratively until approaching an optimum. In EDACE, a matrix called a correlation between binary elements (CBE) matrix is used to represent a binary population. The matrix can be denoted as , where the value of the element indicates the correlation between element and element of a binary design solution. The higher value of means the higher probability that binary elements and will have the same value. The algorithm is developed to deal with a box-constrained optimisation problem:where is an objective function and is a vector containing design variables (a design vector). and are the lower and upper bounds of , respectively. Assuming that a design vector can be represented by a row vector of binary bits size , the CBE matrix thus has the size of . It should be noted that the details of converting a binary string to be a design vector can be found in [55]. In generating a binary string from the CBE matrix, a reference binary solution (RBS) is needed. It can be a randomly generated solution or the best solution found so far depending on a user preference. Then, a row of the matrix is randomly selected (say the th row). The th element of a generated binary solution is set to be the th element of the reference binary solution. The rest of the created binary elements are based on the value of ; . The procedure for creating a binary solution sized from the CBE matrix is detailed in Algorithm 1 where is a binary design solution, is the reference binary solution, is a population size, and is a uniform random number. The algorithm spends loops for creating binary solutions. The process for generating a binary solution from the CBE matrix is in steps –. For one binary solution, only one randomly selected row of CBE (say row ) is used (step ). Then, the th element of a generated binary solution is set equal to the th element of the reference binary solution, . The rest of the elements of the generated binary solution are created in such a way that their values depend on corresponding elements on the th row of CBE. From the computation steps –, the value of determines the probability of to be the same as . The higher value of means the higher correlation between elements and and consequently the higher probability that will be set equal to .

Input: ,
Output: for
Main procedure
Set .
For to
Set a vector used to contain elements of a generated binary string.
Randomly select a position (th row) of .
Set . % Set the th element of a as the th element of .
For
If
% and values are equal, which are either “0” or “1”.
Else
% If , or vice versa.
End
End
Set .
End

The CBE matrix is a square symmetric matrix with equal size to the length of a binary solution whose all diagonal elements are equal to one. For an iteration, the matrix will be updated according to the so far best solution (). The learning rate () with be used to control the changes in updating as with PBIL. Once is updated, the value of is set to be which means the process requires updates since is always set to be 1. The updated denoted by can be calculated fromwhere is the learning rate randomly generated in the interval . and are the th and th elements of , respectively. From the updating equation, if the th and th elements are similar, it means they are correlated; consequently, the value of (and ) is increased. If they are dissimilar or uncorrelated, is then decreased. Nevertheless, the value of must be limited to the predefined interval where and are the predefined lower and upper limits of . Equation (3) is used to maintain diversity in optimisation search. In the original PBIL, a mutation operator is used with the same purpose. Therefore, the procedure of EDACE starts with an initial matrix for correlation between binary elements where and . This implies that when generating a binary solution, its elements have equal probability to be 1 or 0 where its th element can be 1 or 0, created at random. The procedure for general purpose of EDACE is given in Algorithm 2. The decision on selecting for generating a binary solution and for updating the CBE matrix is dependent on a preference of a user. This means other versions of EDACE can be developed in the future.

Input: number of generation (), population size (), binary length ()
Output: ,
Initialisation:
(0.1) Assign and , sized .
(0.2) Randomly generate binary solutions and decode them to be .
(0.3) Calculate objective function values where fun is an objective function evaluation.
(0.4) Find , ,
Main iterations
For to
Update using Equation (2)
Generate from using Algorithm 1, and decode them to be .
For to
Calculate objective function values .
If
, ,
End
End
Update , ,
End

An initial binary population is randomly created. The binary solutions are then decoded to be real design variables where function evaluations are performed and and are found. Then, new binary solutions are generated using Algorithm 1 while the greedy selection (steps –) is activated with and being determined. The CBE matrix is updated by using as detailed in (2)-(3). The search process is repeated until termination criterion is reached. The generation of a binary design solution of EDACE is, to some extent, similar to those used in binary PSO [49] and binary quantum-inspired gravitational search algorithm (BQIGSA) [52] in the sense that the binary solution is controlled by the probability of being “1” or “0”. However, in EDACE, a generated solution relies not only on such probability but also on the reference binary solution . Apart from that, the update of CBE tends to be similar to the concept employed in PBIL with a learning rate and this is totally different from binary PSO and BQIGSA.

In selecting and , if both solutions are the same which is , it could lead to a premature convergence. If both are set to be a solution randomly selected solution from the current binary population, the diversification increases but the convergence rate will be slower. Therefore, the balance between intensification and diversification must be made. In this work, the so far best binary solution is set to be to maintain intensification. For updating the CBE matrix, we use the new updating scheme asThe solutions and are two types of best solutions. Firstly, best solutions are selected from (see Algorithm 2 for both solution sets), sorted according to their functions, and then saved to a set Best_sol. Four vectors are created as the so far best solution, a solution whose elements are averaged from the elements of the first (default = 10) best solutions found so far, a solution whose elements are averaged from the elements of the members of Best_sol, and a solution whose elements are averaged from the elements of the current binary population. is randomly chosen from the aforementioned solutions (, , , and ) with equal probability while is randomly chosen from the members of Best_sol. With this idea, the balance between exploration and exploitation is maintained throughout the search process. Algorithm 3 shows the new CBE updating strategy.

Input: , , , , , Best_sol,
Output:
Main procedure
Create , , ,
For to
Assign .
If , set
If , set
If , set
Otherwise, set
Random selected a vector from Best_sol.
For to
Generate .
Update using Equation (4).
Limit to the interval .
End
End

3. Experimental Set-Up

To investigate the search performance of the proposed algorithm, fifteen learning-based test problems from CEC2015 and one flight dynamic control optimisation problem are used. The former are used for testing the performance of EDACE for general types of box-constrained optimisation while the latter is the real-world application.

3.1. CEC2015 Learning-Based Test Problems

The CEC2015 learning-based test problems are box-constrained single objective benchmark functions proposed in [56]. The problems consist of 2 Unimodal Functions, 3 Simple Multimodal Functions, 3 Hybrid Functions, and 7 Composition Functions. The summary of CEC2015 learning-based test problems is shown in Table 1. It should be noted that the details and the codes for the test problems can be downloaded from the website of CEC2015 competition.

3.2. Flight Dynamic Control Optimisation Problem

Flight dynamic control system design is a classical important application for real engineering problems. The motion of an aircraft can be described using the body axes which is herein the stability axes consisting of roll axis (), pitch axis (), and yaw axis () as shown in Figure 1. The motion of the aircraft is described by Newton’s 2nd law or equations of motion for both translational and rotational motions. The dynamical model is nonlinear but can be linearised by applying aerodynamic derivatives. Due to aircraft symmetry with respect to the plane, the linearised dynamical model can be decoupled into two groups as longitudinal motion and the lateral/directional motion. For more details of deriving the equations of motion, see [57]. In this work, only the lateral/directional motion control is considered. A state equation representing the dynamic motion of an aircraft is expressed as follows [57–60]:where , is the sideslip, a velocity in direction, is the yaw rate, rate of change of rotation about the -axis, is the roll rate, rate of change of rotation about the -axis, is the bank angle, rotation about the -axis, is the kinetic energy matrix, is Coriolis matrix, is the control vector, is the aileron deflection, and is the rudder deflection.

The control vector can be expressed aswhere is a pilot’s control input vector while and are the gain matrices expressed as follows [59]:where parameters are control gain coefficients which need to be found.

From (5)-(6), the state equation for lateral/directional motion of an aircraft can be expressed as

Design optimisation of the control system of an aircraft is found to have many objectives as there are several criteria that need to be satisfied such as control stability, accuracy, sensitivity, and control effort, while the control gains coefficients are set to be design variables for an optimisation problem. In this work, the optimal flight control of an aircraft focuses on only the stability aspect. The objective function is posed to minimise spiral root subjected to stability performance constraints. The optimisation problem can then be written aswhere , , , and are spiral root, roll damping, damping ratio of Dutch-roll complex pair, and Dutch-roll frequency, respectively. These parameters can be calculated based on the eigenvalues associated with the matrix . The design variables are control gain coefficients in the matrix . The kinetic energy matrix () and the Coriolis matrix () are defined as

More details about this aircraft dynamic model can be found in [58–60]. To handle the constraints, the penalty function which was presented in [61] is used.

The proposed EDACE and several well-established binary-code metaheuristics are used to solve the fifteen CEC2015 learning-based test problems and the flight dynamic control test problem. The metaheuristic optimisers are as follows: Genetic algorithm (GA) [21] used binary codes with crossover and mutation rates are 1 and 0.1, respectively. Binary simulated annealing (BSA) [50] used binary codes with exponentially decreasing temperature. The starting and ending temperature are set to be 10 and 0.001, respectively. The cooling step is set as 10. Population-based incremental learning (PBIL) [24] used binary codes with the learning rate, mutation shift, and mutation rate as 0.5, 0.7, and 0.2, respectively. Binary particle swarm optimisation (BPSO) [49] used binary codes with V-shaped transfer function while the transfer function used is the V-shaped version 4 (V4) as reported in [49]. It is noted that this version is said to be the most efficient version based on the results obtained in [49]. Univariate marginal distribution algorithm (UMDA) [48] used binary codes. The first 20 best binary solutions are used to update the probability matrix. Estimation of distribution algorithm with correlation of binary elements (EDACE) (Algorithm 2) used binary codes with , , , , and .

Each algorithm is used to solve the problems for 30 optimisation runs. The population sizes are set to be 100 and 20 while number of generation is set to be 100 and 500 for the CEC2015 learning-based test problems and the flight dynamic control test problem, respectively. For an algorithm using different population size and number of generations such as BSA, it will be terminated at the same number function evaluations, which is 10,000 for all test problems. The binary length is set to be 5 for each design variable for all optimisers.

4. Optimum Results

4.1. CEC2015

After applying the proposed EDACE and several well-established binary MHs for solving the CEC2015 learning-based benchmark functions, the results are shown in Tables 2–4. Note that, apart from the algorithms used in this study, the results of solving CEC2015 test suit obtained from efficient binary artificial bee colony algorithm based on genetic operator (GBABC), binary quantum-inspired gravitational search algorithm (BQIGSA), and self-adaptive binary variant of a differential evolution algorithm (SabDE) as reported in [53] are also included in the comparison. From Table 2, the mean (Mean) and standard deviation (STD) values of the objective functions are used to measure the search convergence and consistency of the algorithms. The lower Mean is the better convergence while the lower STD is the better consistency. The value of Mean is more important; thus, for method A with lower Mean but higher STD than method B, method A is considered to be superior.

For the measure of search convergence based on the mean objective function values, the best performer for the unimodal test functions, and , is EDACE while the second best is BPSO. For the simple multimodal functions, the best performer for and is SabDE while the best performer for is BPSO. The second best performers for , , and are SabDE, BEDACE, and UMDE, respectively. For the hybrid functions, the best performers for the functions , , and , are SabDE, EDACE, and BPSO, respectively, while the second best performer for and is BPSO and the second best for is EDACE. For the final group of CEC2015 test problems, composition functions, the best performer for the , , and is SabDE while the best performers for the and are BPSO and EDACE, respectively. For , the best performers are UMDA, BPSO, GA, PBIL, and EDACE, which obtain the same mean values while, for , the best performers are UMDA, BPSO, GA, PBIL, BSA, and EDACE, which obtained the same mean values. It should be noted that the results from [53] were obtained from using the total number of function evaluations as 1,000,000 with the binary length of 50 for each design variable whereas this work uses 10,000 function evaluations with the binary length of 5 for each design variable. This indirect comparison with GBABC, BQIGSA, and SabDE can only be used to show that the proposed EDACE also has good performance and cannot be used to claim which method is superior.

For the measure of search consistency based on the STD values, the most consistent methods for unimodal functions, and , are BPSO and EDACE while the second most consistent methods are EDACE and BPSO, respectively. For the simple multimodal functions, the best for and is SabDE while the best for is the proposed EDACE. EDACE is the best for the hybrid function of while BPSO is the best for the hybrid functions and . For the composition functions, EDACE is the best for the problems and while BPSO is the best for . For the composition functions, , , and , the best is SabDE while the best for is BSA.

The value Min in Table 2 is the objective function value of the best run from a particular method. Note that only the UMDA, BPSO, GA, PBIL, BSA, and EDACE were compared. For the unimodal function, the minimum objective function values of and were obtained by BPSO and EDACE, respectively. For the simple multimodal functions, the minimum objective function values for and are obtained from BPSO and EDACE, respectively, while for , the minimum is obtained from UMDA, BSA, and EDACE. The EDACE obtained minimum objective function values for all test functions in the hybrid function group. However, for the hybrid function , three algorithms including BPSO, GA, and EDACE obtained the minimum values. For the composition functions, EDACE obtained the minimum function values for all test functions. However, for the functions and , all algorithms obtained the same minimum values while for the , BPSO and EDACE obtained the same minimum function values. Similarly, for , UMDA, BPSO, BSA, and EDACE obtained the same minimum values.

Table 3 shows the summary of ranking based on the mean objective function values from 30 optimisation runs. It was found that the proposed EDACE is mostly ranked in top three best from solving fifteen CEC2015 learning-based test problems. After summing up the ranking score, it is found that EDACE and BPSO are equal best performer while the third best is UMDA.

In order to further investigate the performance comparison of the binary-code MHs, the statistical -test is employed. Table 4 shows a 9 × 9 comparison matrix of the 9 optimisers. If method is significantly better than method based on the -test at 5% significant level, the column and row of the matrix are set to be 1; otherwise, they are set to be 0. When summing up along the columns, the highest score indicates the best optimiser based on this type of comparison. In the table, it means EDACE is the best. Table 5 shows the ranking of the 9 optimisers when solving all CEC2015 learning-based test problems based on the -test. After summing up the ranking numbers of all test problems, it is found that EDACE is the overall best optimiser while BPSO and UMDA are the second and the third best, respectively.

Figures 2–5 show the search history of the top three optimisers EDACE, BPSO, and UMDA on solving all CEC2015 learning-based test problems where the vertical axis is the average objective function from 30 runs of each method. For all test functions, it was found that EDACE and UMDA converged to the optimal values at higher speed while BPSO seems to converge slowly and consistently. However, for all functions, BPSO finally moves to the minimum or near minimum function values at the end of the runs. EDACE shows fast convergence from the beginning and obtained the minimum or near minimum values for all test functions except for . This indicates the ability of search exploitation and search exploration of the proposed EDACE since the CEC2015 test functions were assigned to test both aspects of MHs.

Table 6 shows performance of EDACE on solving unimodal function, , when the binary lengths for each design variable are 5, 10, 25, and 50 for 10 optimisation runs. It was found that when the number of binary bits increases, the computational time increases and the resulting mean objective function values decrease for the binary lengths less than 25. However, for the binary length of 50, the mean objective function value increases meaning EDACE performance deteriorates. Without considering computational time, the best number of binary length is 25.

4.2. Flight Dynamic Control System Design

After applying the six binary-code MHs to solve the real engineering application of flight dynamic and control system for 30 optimisation runs, the comparison results are shown as box-plots of the objective and constraint violation values (Figure 6). The upper and lower horizontal lines of each box represent the maximum and minimum of objective function values, respectively, while the internal line shows the median of objective function values. From this figure, based on median values of objective function, it is found that the best performer is EDACE while the second best and the third best are BPSO and UMDA, respectively. The most consistent method having the smallest gap between the maximum and minimum for all of optimisation runs is UMDA. However, the worst function value found by EDACE is almost as good as the best found by UMDA. Thus, the proposed EDACE is superior. Based on the figure, it was found that GA failed to solve the problem as it cannot obtain a feasible optimum point. The minimum objective function value is obtained from using the proposed EDACE.

Figure 7 shows the best run search history of all optimisers (selection based on the minimum objective function values of feasible solutions). From the figure, UMDA and PBIL seem to be the fastest convergent methods initially. However, after the process goes on for about 4,000 function evaluations, the proposed EDACE converged to the minimum objective function value with a faster rate than the others. It has better exploration rate as the best function value is still decreased at the late iteration numbers. BPSO, on the other hand, seems to be slower than UMDA, PBIL, and BSA in the beginning. It however can converge to the better results after around 8,000 function evaluations.

5. Conclusions and Discussion

In this work, a new concept of a binary-code optimiser is proposed. Fifteen CEC2015 learning-based test problems and a real engineering design problem of flight dynamic and control system are used to investigate the search performance of the proposed algorithm. Several well-established binary-code MHs are used in comparison. The results obtained show that the proposed EDACE is the best performer on solving the 15 CEC2015 learning-based test problems and real engineering design problem of flight dynamic and control. Further improvement of EDACE by means of self-adaptation will be investigated in the future. The choice for needs further studies. The use of EDACE for hyperheuristic development is also possible. The extension to multiobjective optimisation and many-objective optimisation is also under investigation. Appling EDACE for the more complex problems such as large scale problems, mixed-variable problems, and reliability optimisation is for future work. The fight control optimisation problem, one of our recent research focuses, has more than three objective functions to be optimised; thus, it should be formulated as many-objective optimisation. This along with aircraft path planning dynamic optimisation still needs considerably more investigation while EDACE will be one of optimisers to be used for solving such design problems.

Conflicts of Interest

The authors declare that there are no conflicts of interest regarding the publication of this article.

Acknowledgments

The authors are grateful for the support from the Thailand Research Fund (TRF), Grant no. MRG5980238.

References

S. Mirjalili, “Moth-flame optimization algorithm: a novel nature-inspired heuristic paradigm,” Knowledge-Based Systems, vol. 89, pp. 228–249, 2015.
View at: Publisher Site | Google Scholar
R. V. Rao, V. J. Savsani, and D. P. Vakharia, “Teaching-learning-based optimization: a novel method for constrained mechanical design optimization problems,” Computer-Aided Design, vol. 43, no. 3, pp. 303–315, 2011.
View at: Publisher Site | Google Scholar
Y. Tan and Y. Zhu, “Fireworks algorithm for optimization,” in Advances in Swarm Intelligence, Y. Tan, Y. Shi, and K. Tan, Eds., vol. 6145, pp. 355–364, Springer, Berlin Heidelberg, Germany, 2010.
View at: Publisher Site | Google Scholar
A. Kaveh and S. Talatahari, “A novel heuristic optimization method: charged system search,” Acta Mechanica, vol. 213, no. 3-4, pp. 267–289, 2010.
View at: Publisher Site | Google Scholar
E. Rashedi, H. Nezamabadi-pour, and S. Saryazdi, “GSA: a gravitational search algorithm,” Information Sciences, vol. 179, no. 13, pp. 2232–2248, 2009.
View at: Publisher Site | Google Scholar
M. Fesanghary, “Harmony search applications in mechanical, chemical and electrical engineering,” in Music-Inspired Harmony Search Algorithm, Z. Geem, Ed., vol. 191, pp. 71–86, Springer Publishing Company, Berlin, Heidelberg, Germany, 2009.
View at: Publisher Site | Google Scholar
X. Wu, Y. Zhou, and Y. Lu, “Elite opposition-based water wave optimization algorithm for global optimization,” Mathematical Problems in Engineering, vol. 2017, Article ID 3498363, 25 pages, 2017.
View at: Publisher Site | Google Scholar
C. Wang, Y. Wang, K. Wang, Y. Dong, and Y. Yang, “An improved hybrid algorithm based on biogeography/complex and Metropolis for many-objective optimization,” Mathematical Problems in Engineering, vol. 2017, Article ID 2462891, 14 pages, 2017.
View at: Publisher Site | Google Scholar | MathSciNet
W. Lei, H. Manier, M.-A. Manier, and X. Wang, “A hybrid quantum evolutionary algorithm with improved decoding scheme for a robotic flow shop scheduling problem,” Mathematical Problems in Engineering, vol. 2017, Article ID 3064724, 13 pages, 2017.
View at: Publisher Site | Google Scholar | MathSciNet
C.-R. Hwang, “Simulated annealing: theory and applications,” Acta Applicandae Mathematicae, vol. 12, pp. 108–111, 1988.
View at: Google Scholar
C. Qu, S. Zhao, Y. Fu, and W. He, “Chicken swarm optimization based on elite opposition-based learning,” Mathematical Problems in Engineering, vol. 2017, Article ID 2734362, 20 pages, 2017.
View at: Publisher Site | Google Scholar | MathSciNet
N. Dong, X. Fang, and A.-g. Wu, “A novel chaotic particle swarm optimization algorithm for parking space guidance,” Mathematical Problems in Engineering, vol. 2016, Article ID 5126808, 14 pages, 2016.
View at: Publisher Site | Google Scholar | MathSciNet
X.-S. Yang and A. H. Gandomi, “Bat algorithm: a novel approach for global engineering optimization,” Engineering Computations, vol. 29, no. 5, pp. 464–483, 2012.
View at: Publisher Site | Google Scholar
X.-S. Yang and S. Deb, “Engineering optimisation by Cuckoo search,” International Journal of Mathematical Modelling and Numerical Optimisation, vol. 1, no. 4, pp. 330–343, 2010.
View at: Publisher Site | Google Scholar
K. Socha and M. Dorigo, “Ant colony optimization for continuous domains,” European Journal of Operational Research, vol. 185, no. 3, pp. 1155–1173, 2008.
View at: Publisher Site | Google Scholar | MathSciNet
D. Karaboga and B. Basturk, “A powerful and efficient algorithm for numerical function optimization: artificial bee colony (ABC) algorithm,” Journal of Global Optimization, vol. 39, no. 3, pp. 459–471, 2007.
View at: Publisher Site | Google Scholar | MathSciNet
G. Venter and J. Sobieszczanski-Sobieski, “Particle swarm optimization,” AIAA Journal, vol. 41, no. 8, pp. 1583–1589, 2003.
View at: Publisher Site | Google Scholar
V. Muthiah-Nakarajan and M. M. Noel, “Galactic swarm optimization: a new global optimization metaheuristic inspired by galactic motion,” Applied Soft Computing Journal, vol. 38, pp. 771–787, 2016.
View at: Publisher Site | Google Scholar
S. Mirjalili and A. Lewis, “The whale optimization algorithm,” Advances in Engineering Software, vol. 95, pp. 51–67, 2016.
View at: Publisher Site | Google Scholar
S. Mirjalili, S. M. Mirjalili, and A. Lewis, “Grey wolf optimizer,” Advances in Engineering Software, vol. 69, pp. 46–61, 2014.
View at: Publisher Site | Google Scholar
D. E. Goldberg and J. H. Holland, “Genetic algorithms and machine learning,” Machine Learning, vol. 3, no. 2-3, pp. 95–99, 1998.
View at: Publisher Site | Google Scholar
R. Storn and K. Price, “Differential evolution—a simple and efficient heuristic for global optimization over continuous spaces,” Journal of Global Optimization, vol. 11, no. 4, pp. 341–359, 1997.
View at: Publisher Site | Google Scholar | MathSciNet
T. Bäck, Evolutionary Algorithms in Theory and Practice, Oxford University Press, New York, NY, USA, 1996.
View at: Zentralblatt MATH | MathSciNet
S. Baluja, Population-Based Incremental Learning: A Method for Integrating Genetic Search Based Function Optimization and Competitive Learning, Carnagie Mellon University, Pittsburgh, PA, USA, 1994.
L. J. Fogel, A. J. Owens, and M. J. Walsh, Artificial Intelligence through Simulated Evolution, John Wiley, New York, NY, USA, 1966.
View at: Publisher Site
H. Beyer and H. Schwefel, “Evolution strategies—a comprehensive introduction,” Natural Computing, vol. 1, no. 1, pp. 3–52, 2002.
View at: Publisher Site | Google Scholar | MathSciNet
S. Mirjalili, “SCA: a sine cosine algorithm for solving optimization problems,” Knowledge-Based Systems, vol. 96, pp. 120–133, 2016.
View at: Publisher Site | Google Scholar
R. V. Rao, “Jaya: a simple and new optimization algorithm for solving constrained and unconstrained optimization problems,” International Journal of Industrial Engineering Computations, vol. 7, no. 1, pp. 19–34, 2016.
View at: Publisher Site | Google Scholar
J. Q. Zhang and A. C. Sanderson, “JADE: adaptive differential evolution with optional external archive,” IEEE Transactions on Evolutionary Computation, vol. 13, no. 5, pp. 945–958, 2009.
View at: Publisher Site | Google Scholar
R. Tanabe and A. Fukunaga, “Evaluating the performance of SHADE on CEC 2013 benchmark problems,” in Proceedings of the 2013 IEEE Congress on Evolutionary Computation (CEC '13), pp. 1952–1959, June 2013.
View at: Publisher Site | Google Scholar
R. Tanabe and A. S. Fukunaga, “Improving the search performance of shade using linear population size reduction,” in Proceedings of the IEEE Congress on Evolutionary Computation (CEC '14), pp. 1658–1665, IEEE, July 2014.
View at: Publisher Site | Google Scholar
M. P. Wachowiak, M. C. Timson, and D. J. DuVal, “Adaptive particle swarm optimization with heterogeneous multicore parallelism and GPU acceleration,” IEEE Transactions on Parallel and Distributed Systems, 2017.
View at: Publisher Site | Google Scholar
L. Zhang, Y. Tang, C. Hua, and X. Guan, “A new particle swarm optimization algorithm with adaptive inertia weight based on Bayesian techniques,” Applied Soft Computing Journal, vol. 28, pp. 138–149, 2015.
View at: Publisher Site | Google Scholar
X. Liang, W. Li, Y. Zhang, and M. C. Zhou, “An adaptive particle swarm optimization method based on clustering,” Soft Computing, vol. 19, no. 2, pp. 431–448, 2015.
View at: Publisher Site | Google Scholar
N. Srinivas and K. Deb, “Multiobjective function optimization using nondominated sorting genetic algorithms,” Evolutionary Computation, vol. 2, pp. 221–248, 1994.
View at: Publisher Site | Google Scholar
K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, “A fast and elitist multiobjective genetic algorithm: NSGA-II,” IEEE Transactions on Evolutionary Computation, vol. 6, no. 2, pp. 182–197, 2002.
View at: Publisher Site | Google Scholar
W. Mkaouer, M. Kessentini, S. Bechikh, K. Deb, and M. Ó. Cinnéide, “High dimensional search-based software engineering: Finding tradeoffs among 15 objectives for automating software refactoring using NSGA-III,” in Proceedings of the 16th Genetic and Evolutionary Computation Conference (GECCO '14), pp. 1263–1270, July 2014.
View at: Publisher Site | Google Scholar
C. A. Coello Coello and M. S. Lechuga, “MOPSO: a proposal for multiple objective particle swarm optimization,” in Proceedings of the Congress on Evolutionary Computation (CEC '02), pp. 1051–1056, May 2002.
View at: Publisher Site | Google Scholar
E. Zitzler, M. Laumanns, and L. Thiele, “SPEA2: improving the strength Pareto evolutionary algorithm for multiobjective optimization, presented at the Evolutionary Methods for Design, Optomozation and Control,” Evolutionary Methods for Design, Optomozation and Control, Barcelona Spain, 2002.
View at: Google Scholar
K. Nuaekaew, P. Artrit, N. Pholdee, and S. Bureerat, “Optimal reactive power dispatch problem using a two-archive multi-objective grey wolf optimizer,” Expert Systems with Applications, vol. 87, pp. 79–89, 2017.
View at: Publisher Site | Google Scholar
F. Zou, L. Wang, X. Hei, D. Chen, and B. Wang, “Multi-objective optimization using teaching-learning-based optimization algorithm,” Engineering Applications of Artificial Intelligence, vol. 26, no. 4, pp. 1291–1300, 2013.
View at: Publisher Site | Google Scholar
Q. Zhang and H. Li, “MOEA/D: a multiobjective evolutionary algorithm based on decomposition,” IEEE Transactions on Evolutionary Computation, vol. 11, no. 6, pp. 712–731, 2007.
View at: Publisher Site | Google Scholar
D. Angus and C. Woodward, “Multiple objective ant colony optimisation,” Swarm Intelligence, vol. 3, no. 1, pp. 69–85, 2009.
View at: Publisher Site | Google Scholar
B. V. Babu and M. M. L. Jehan, “Differential evolution for multi-objective optimization,” in Proceedings of the Congress on Evolutionary Computation (CEC '03), pp. 2696–2703, Canberra, Australia, December 2003.
View at: Publisher Site | Google Scholar
X. Zhang, Y. Tian, and Y. Jin, “A knee point driven evolutionary algorithm for many-objective optimization,” IEEE Transactions on Evolutionary Computation, vol. 19, no. 6, pp. 761–776, 2015.
View at: Publisher Site | Google Scholar
H. Wang, L. Jiao, and X. Yao, “Two_Arch2: an improved two-archive algorithm for many-objective optimization,” IEEE Transactions on Evolutionary Computation, vol. 19, no. 4, pp. 524–541, 2015.
View at: Publisher Site | Google Scholar
R. Wang, R. C. Purshouse, and P. J. Fleming, “Preference-inspired coevolutionary algorithms for many-objective optimization,” IEEE Transactions on Evolutionary Computation, vol. 17, no. 4, pp. 474–494, 2013.
View at: Publisher Site | Google Scholar
H. Mühlenbein, “The equation for response to selection and its use for prediction,” Evolutionary Computation, vol. 5, no. 3, pp. 303–346, 1997.
View at: Publisher Site | Google Scholar
S. Mirjalili and A. Lewis, “S-shaped versus V-shaped transfer functions for binary particle swarm optimization,” Swarm and Evolutionary Computation, vol. 9, pp. 1–14, 2013.
View at: Publisher Site | Google Scholar
S. Bureerat and J. Limtragool, “Structural topology optimisation using simulated annealing with multiresolution design variables,” Finite Elements in Analysis and Design, vol. 44, no. 12-13, pp. 738–747, 2008.
View at: Publisher Site | Google Scholar
C. Ozturk, E. Hancer, and D. Karaboga, “A novel binary artificial bee colony algorithm based on genetic operators,” Information Sciences, vol. 297, pp. 154–170, 2015.
View at: Publisher Site | Google Scholar | MathSciNet
H. Nezamabadi-Pour, “A quantum-inspired gravitational search algorithm for binary encoded optimization problems,” Engineering Applications of Artificial Intelligence, vol. 40, pp. 62–75, 2015.
View at: Publisher Site | Google Scholar
A. Banitalebi, M. I. A. Aziz, and Z. A. Aziz, “A self-adaptive binary differential evolution algorithm for large scale binary optimization problems,” Information Sciences, vol. 367-368, pp. 487–511, 2016.
View at: Publisher Site | Google Scholar
B. Crawford, R. Soto, G. Astorga, J. Garcia, C. Castro, and F. Paredes, “Putting continuous metaheuristics to work in binary search spaces,” Complexity, vol. 2017, Article ID 8404231, 19 pages, 2017.
View at: Publisher Site | Google Scholar | MathSciNet
G. Lindfield and J. Penny, Numerical Methods: Using MATLAB, Academic Press, 2012.
J. Liang, B. Qu, P. Suganthan, and Q. Chen, “Problem definitions and evaluation criteria for the CEC 2015 competition on learning-based real-parameter single objective optimization,” Tech. Rep., Computational Intelligence Laboratory, Zhengzhou University, Zhengzhou China and Technical Report, Nanyang Technological University, Singapore, 2014.
View at: Google Scholar
D. A. Caughey, Introduction to Aircraft Stability and Control Course Notes for M&AE 5070, Sibley School of Mechanical & Aerospace Engineering, Cornell University, Ithaca, New York, NY, USA, 2011.
S. Rostami and F. Neri, “Covariance matrix adaptation pareto archived evolution strategy with hypervolume-sorted adaptive grid algorithm,” Integrated Computer-Aided Engineering, vol. 23, no. 4, pp. 313–329, 2016.
View at: Publisher Site | Google Scholar
S. F. Adra, A. I. Hamody, I. Griffin, and P. J. Fleming, “A hybrid multi-objective evolutionary algorithm using an inverse Neural Network for aircraft control system design,” in Proceedings of the 2005 IEEE Congress on Evolutionary Computation (IEEE CEC '05), pp. 1–8, September 2005.
View at: Google Scholar
D. Tabak, A. A. Schy, D. P. Giesy, and K. G. Johnson, “Application of multiobjective optimization in aircraft control systems design,” Automatica, vol. 15, no. 5, pp. 595–600, 1979.
View at: Publisher Site | Google Scholar
S. Bureerat and N. Pholdee, “Optimal truss sizing using an adaptive differential evolution algorithm,” Journal of Computing in Civil Engineering, vol. 30, no. 2, p. 04015019, 2015.
View at: Publisher Site | Google Scholar

Copyright

Copyright © 2017 Nantiwat Pholdee and Sujin Bureerat. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies

Views

1111

Downloads

664

Citations